ssadqdsacf/cross-unlearning-case6-qwen35-4b-ga-epoch10

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 23, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The ssadqdsacf/cross-unlearning-case6-qwen35-4b-ga-epoch10 is a 4.5 billion parameter Qwen3.5-based model, specifically a 'ga unlearning baseline' initialized from an epoch-10 SFT model. This model is one of six formal matrix models, designed for cross-unlearning research and includes verified three-direction classification, mask, and generation summaries. It is intended for evaluating unlearning protocols and model behavior under specific training conditions.

Loading preview...

Model Overview

The ssadqdsacf/cross-unlearning-case6-qwen35-4b-ga-epoch10 is a 4.5 billion parameter model based on the Qwen3.5 architecture. It serves as a 'ga unlearning baseline,' initialized from a selected epoch-10 SFT (Supervised Fine-Tuning) model. This model is part of a formal matrix of six models, focusing on research into cross-unlearning protocols.

Key Characteristics

  • Architecture: Qwen3.5-4B base model.
  • Parameter Count: 4.5 billion parameters.
  • Context Length: Supports a context length of 32768 tokens.
  • Unlearning Baseline: Functions as a baseline for 'ga unlearning' research, derived from an SFT model at epoch 10.
  • Provenance: Detailed provenance is available, including protocol (qwen3.5-4b-profile-400-sft-v2-r32-cosine-15e-selected-epoch10), SHA256 hashes, and model fingerprints.
  • Evaluation: Evaluated using fixed 4-shot greedy decoding, with classification employing native thinking and mask/generation disabling thinking.
  • Included Data: The repository provides a directly loadable merged model, a LoRA adapter, and verified three-direction classification, mask, and generation summaries.

Intended Use Cases

This model is primarily designed for:

  • Research in Cross-Unlearning: Investigating and evaluating unlearning protocols and their effects on model behavior.
  • Protocol Analysis: Analyzing the impact of specific training and selection protocols on model performance and characteristics.
  • Comparative Studies: Serving as a controlled baseline within a matrix of models for comparative research.

Finetunes
1 models